MPGreen

Raw data - example

Inspiration

Urban areas are increasingly polluted by traffic, and although people make an effort to ride share, cars still produce a large portion of urban carbon dioxide emissions. App-based directions offer an option between distance-optimized or time-optimized routes, but never present the possibility of an eco-friendly route. In addition, many people want to be more green, but don't know what factors most impact the carbon footprint of their vehicle.

What it does

Our interface provides an information-based solution. By synthesizing millions of very precise data points from Ford's OpenXC platform, we can isolate factors like idling, tailgating, aggressive driving, and analyze their impacts on fuel efficiency. We strip the noise in raw data to find the important trends and present them in clear visualizations.

How we built it

The bulk of data analysis is handled by python which processes the raw json data files. Then, we use pandas to streamline data into tables, which are easier to handle and filter. Given the extreme precision of the data points (records in fractions of a second), the data was initially very difficult to interpret. With the help of numpy, we were able to efficiently calculate MPG figures and overlay additional trends on several visuals.

Challenges we ran into

Data points for specific vehicle attributes are taken very irregularly and do not match up at the same timestamps. The user's interaction with their car's usage - negative fuel usage figures when tanks were filled. Column names in the data were inconsistent across sets (e.g. Odometer vs Fine Odometer Since Restart). Plenty of files had missing data for certain attributes, resulting in a scattering of NaNs across the dataset. Given this, we had to be clever with data filtering and condense the data so important metrics could be compared.

Accomplishments that we're proud of

Beautiful visuals indicating clear trends in data. Clean filtering of extremely noisy raw data. A fun frontend that's visually appealing to the user.

What we learned

Big data is not as easy as running a few functions on data that's simply downloaded from a database. Much of analytics is the filtration and data handling, and trends may often be surprising.

What's next for MPGreen

We could integrate Maps and Directions APIs to find more eco friendly routes in order to directly provide the user with ways to reduce their carbon footprint. As it stands, our system is a strong tool to view and share information, but has potential to actually impact the environment.